Voice wake-up interaction response method and system

Abstract

The invention discloses a voice wake-up interaction response method and system, which are mainly used for detecting whether a user calls after a voice interaction system is awakened. Dividing the wakened given time window into different areas, and processing the different areas by adopting different technical means; the method comprises the following steps: detecting whether a user call exists in a false triggering area of an adjacent time area after wake-up word detection, and judging whether the call is confused with a tail sound of the wake-up word; detecting whether there is a user call in a normal call detection area after the false triggering area; and detecting whether there is a user call in a blind area detection area after the normal call detection area, under the condition that no call content is detected in the blind area detection area, carrying out fundamental frequency calculation on the blind area detection area, and if a call with a certain duration can be obtained through fundamental frequency calculation, considering that the call is detected. By adopting the method and the system provided by the invention, the proportion of false detection and missing detection can be greatly reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent voice interaction, and in particular relates to an interactive response method and system for voice wake-up. Background Art

[0002] With the development of society and electronic information technology, artificial intelligence products have become an indispensable necessity in people's lives, such as smart speakers, smart cars, smart TVs, smart air conditioners, etc. Many other devices and applications are also rapidly developing in the direction of intelligence. At the same time, users' functional requirements for artificial intelligence products are also rich and diverse. In addition to meeting basic functions, they also require audio-visual entertainment, life services, and interconnection.

[0003] With the enrichment of smart applications and the development of AI technology, consumers' understanding of smart products is constantly upgrading, and the product experience is also changing from a single function to more smart scenarios. Therefore, many functions of products need to be redefined. At the same time, the substantial progress of technologies such as artificial intelligence, 5G, human-computer interaction devices and operating systems is driving the rapid development of smart applications to meet the growing awareness and needs of users.

[0004] In the process of interaction between smart voice devices and people, voice wake-up is a basic capability used to start interaction with smart devices. The purpose of voice wake-up is to activate smart devices from dormant state to running state, so after the wake-up word is spoken, it can be detected immediately, and the user experience will be better. For example, Apple's "Hey, Siri" and Baidu's "Xiaodu Xiaodu" are both existing wake-up words.

[0005] Generally, after the user wakes up, the system will give the user a response, such as "I'm here". However, if the user speaks immediately after waking up, the user's words may overlap with the system's announcement, thus affecting the interactive experience and effect.

[0006] In the prior art, there is a form of interaction in which the system detects whether the user has spoken a voice command immediately after the user wakes up with voice, which is also called "one-time" interaction. After waking up, the system will detect whether the user has spoken immediately afterwards, and use this as a basis to decide whether to broadcast a welcome message. If the system detects that the user has spoken immediately, it will not broadcast a response message such as a welcome message to avoid interfering with the user's input and affecting the experience and effect. The above-mentioned voice wake-up interaction process in the prior art is shown in the attached figure of the specification. Figure 1 shown.

[0007] According to the attached drawings of the specification Figure 1In the prior art, a common practice is to detect whether the user has spoken within a given time window (such as 400ms) after wake-up through voice endpoint detection. This may lead to two problems: first, the process of detecting speech is easily affected by the tail sound of the wake-up word; second, the process of detecting speech is easy to miss detection in the second half of 400ms. This makes the detection effect poor and prone to missed detection or false detection. Summary of the invention

[0008] In order to solve the above defects in the prior art, the present invention proposes a new method for detecting whether the user has spoken after being awakened, dividing the given time window after awakening into different areas, and using different technical means to process in different areas to improve the interactive response of voice wake-up.

[0009] The present invention provides an interactive response method for voice wake-up, comprising the following steps:

[0010] Get user's speech;

[0011] Detecting whether the user's speech contains a wake-up word;

[0012] After the wake-up word is detected, it is detected whether there is a user speaking in the detection area after the wake-up word;

[0013] Wherein, the detection within the detection area includes detecting whether a user speaks within a normal speaking detection area;

[0014] If no user speech is detected in the detection area, a welcome message is played;

[0015] The detection within the detection area also includes:

[0016] Detecting whether a user speaks in the false trigger area in the time area immediately after the wake-up word is detected, and if a speech is detected, further determining whether the speech is confused with the tail sound of the wake-up word; if there is confusion, it is considered that no speech is detected;

[0017] and / or

[0018] The blind spot detection area behind the normal speech detection area detects whether the user speaks. If no speech content is detected in the blind spot detection area, the fundamental frequency is calculated for the blind spot detection area. If the fundamental frequency calculation can obtain speech with a certain duration, it is considered that the speech is detected.

[0019] Furthermore, a wake-up model is used to detect the user speaking in the detection area.

[0020] Furthermore, a correlation analysis method is used to determine whether the utterance is confused with the ending sound of the wake-up word.

[0021] The present invention provides an interactive response system for voice wake-up, the system comprising:

[0022] The acquisition module is used to obtain the user's speech;

[0023] A wake-up word detection module, used to detect whether there is a wake-up word in the user speech acquired by the acquisition unit;

[0024] A user speech detection module is used to detect the second speech of the user after the wake-up word;

[0025] Wherein, the user speech detection module includes a normal speech detection module, and includes a false trigger detection module and / or a blind spot detection module;

[0026] The false trigger detection module is used to detect whether the second utterance exists within a period of time immediately after the wake-up word, and if the second utterance is detected, it is necessary to further determine whether the second utterance is confused with the tail sound of the wake-up word;

[0027] The normal speech detection module is used to detect whether the second speech exists within a period of time after the wake-up word or the false trigger detection;

[0028] The blind spot detection module is used to detect whether the second speech exists within a period of time after the normal speech detection, and if the second speech is still not detected, perform baseband detection in the time period to determine whether the user speaks;

[0029] The voice output module is used to output a welcome message when the user speech detection module does not detect the second speech after the system wakes up.

[0030] Furthermore, the user speech detection module uses a wake-up model to perform detection when detecting the second speech.

[0031] Furthermore, the false trigger detection module uses a correlation analysis method to determine whether the second utterance is confused with the tail sound of the wake-up word.

[0032] The present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of implementing the method are realized.

[0033] The present invention provides an electronic device, comprising:

[0034] one or more processors; and

[0035] A memory associated with the one or more processors, the memory is used to store program instructions, and when the program instructions are read and executed by the one or more processors, the steps of the method are executed.

[0036] Compared with the prior art, the present invention has the following advantages and positive effects: the method provided by the present invention can effectively improve the problems existing in the prior art, and is verified using actual data from a normal interaction process, with the false detection rate being below 0.01%, and missed detection not occurring in the actual process. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] For a more complete understanding of the present invention, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:

[0038] Figure 1 A flow chart of voice wake-up interaction in the prior art;

[0039] Figure 2 A flow chart of the interactive response method for voice wake-up provided by the present invention;

[0040] Figure 3 A schematic diagram of the division of user speech detection areas of the present invention;

[0041] Figure 4 A schematic diagram of the interactive response system for voice wake-up provided by the present invention. DETAILED DESCRIPTION

[0042] In order to make the problems, technical solutions and technical effects to be solved by the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments and obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0043] Before describing the exemplary embodiments of the present invention in more detail, it should be noted that although the flowchart of the present invention describes the operations as a sequential processing mode, many of the operation steps can be implemented in parallel or simultaneously, and the order of the operations can also be rearranged as needed. In addition, the processing can be terminated when the operation is completed, but there can also be additional steps not included in the drawings.

[0044] In the following description, the terms "first, second, third", etc. are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first, second, third", etc. can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0045] Embodiment 1

[0046] The present invention provides an interactive response method for voice wake-up, as shown in the accompanying drawings of the specification Figure 2 As shown, it mainly includes the following steps:

[0047] Step S01: obtaining a first utterance of a user, and detecting whether the first utterance of the user contains a wake-up word.

[0048] For example, the wake-up word is "Hello Xiaoice"; the user audio is collected, and the similarity between the user audio and the wake-up word "Hello Xiaoice" is detected. If the similarity is greater than the set threshold, it is considered that the user's first utterance contains the wake-up word.

[0049] Step S02: Detect the second speech of the user after the wake-up word in the first speech in the detection area, and divide the detection area into a false trigger area, a normal area, and a blind area, as shown in the attached figure of the specification Figure 3 and perform the following steps:

[0050] Step S201: Detect user speech in the false trigger area after wake-up. If speech is detected, further determine whether the speech is confused with the tail sound of the wake-up word; if there is confusion, it is considered that no speech is detected.

[0051] The false trigger area is the time area immediately after the wake-up word detection area, for example, within a duration of 100ms.

[0052] When detecting speech in the false trigger area, it is necessary to determine whether the current speech is caused by the tail tone of the wake-up word. For example, the wake-up word is "Hello Xiaobing", if the word "bing" is spoken for a long time, the wake-up is triggered before "bing" is finished. In this way, the tail tone of "bing" is easily triggered by mistake and mistaken for a new voice command. Therefore, after detecting the wake-up word, in the false trigger area after wake-up, for example, within a duration of 100ms, it is detected whether the content of the speech is confused with "bing".

[0053] When detecting whether the voice command is confused with the ending sound of the wake-up word, a correlation analysis method may be used to analyze whether the detected content is correlated with the ending sound of the wake-up word.

[0054] When performing correlation analysis, a confusion table that records the confusion or similarity relationship of each utterance content can be used to effectively solve this type of problem.

[0055] Step S202: Detecting a user speaking in the normal area after the false trigger area.

[0056] Step S203: Detect speech in the blind area behind the normal area. If no speech content is detected in the blind area, perform fundamental frequency calculation on the blind area. If the fundamental frequency calculation can obtain speech with a certain duration, it is considered that there is speech after wake-up.

[0057] After the normal area, the blind area is set, such as setting the duration to 120ms. Since the blind area is the last section of the second speech detection, the final result will be given after the detection and processing are completed. Therefore, excessive delay is not allowed in the blind area. In other words, the detection time of the blind area is relatively short, which may cause the user's speech to be unable to be detected in time. In this regard, low-delay technology, such as F0 (base frequency), can be used for auxiliary judgment.

[0058] When a person speaks, the sound is divided into unvoiced and voiced. The voiced signal is generated by the vibration of the vocal cords and has obvious periodicity. The voiced part has a fundamental frequency that can be extracted. The basic principle of fundamental frequency detection is to use the periodicity of the voiced signal to estimate the period of the voiced signal to obtain the fundamental frequency, that is, the frequency of the vocal cord vibration. Extract the fundamental frequency of the current audio and judge whether there is speech based on it. If there is a certain continuous speaking time, such as 50ms, through the fundamental frequency calculation, it is also considered that there is speech after awakening.

[0059] In step S02, the detection area may also be divided into two areas, such as a tail tone detection area (false trigger area) and a real-time speech area (normal area), and corresponding detection steps may be performed.

[0060] Speech content detection generally uses a wake-up model or automatic speech recognition (Automatic Speech Recognition, ASR model). The present invention directly uses the wake-up model as a content detection technology method. In view of the different characteristics of the false trigger area, the normal area, and the blind area, it is shared with other detection methods. For example, the false trigger area has normal interference, and it is necessary to determine whether it is falsely triggered by the wake-up word after detection in conjunction with the detection results of the model; after the blind area is detected using the wake-up model, it is necessary to use low-latency technology for secondary confirmation.

[0061] Step S03, based on the detection results in step S02, determine whether there is a user speaking in the detection area, if no user speaking is detected, play a welcome message; if a user speaking is detected, perform an interactive response based on the user speaking.

[0062] For example, when a user says, "Hello Xiaoyi, please open the car window for me", the detection area after the wake-up word "Hello Xiaoyi" detects the user saying "Hello Xiaoyi" and the welcome message is played based on the user's words. If a user says, "Hello Xiaoyi <pause 600ms> please open the car window for me", the detection area after the wake-up word "Hello Xiaoyi" does not detect the user saying "Hello Xiaoyi" and the welcome message is played.

[0063] Embodiment 2

[0064] Corresponding to the method described in the present invention, the present invention provides an interactive response system for voice wake-up, see Figure 4 , the system comprising:

[0065] The acquisition module 100 is used to acquire the user's speech.

[0066] The wake-up word detection module 200 is used to detect whether there is a wake-up word in the user speech acquired by the acquisition unit 101.

[0067] The user speech detection module 300 is used to detect the second speech of the user after the wake-up word.

[0068] The user speech detection module 300 includes a false trigger detection module 301 , a normal speech detection module 302 , and a blind spot detection module 303 .

[0069] The false trigger detection module 301 is used to detect whether the second utterance exists within a period of time after the wake-up word. If the second utterance is detected, it is necessary to further determine whether the second utterance is confused with the tail sound of the wake-up word.

[0070] The normal speech detection module 302 is used to detect whether the second speech exists within a period of time after the wake-up word or the false trigger detection.

[0071] The blind spot detection module 303 is used to detect whether the second speech occurs within a period of time after the normal speech detection. If the second speech is still not detected, a baseband detection is performed during the period of time to further determine whether the user speaks.

[0072] The voice output module 400 is used to output a welcome message when no user speech is detected after the system is awakened.

[0073] In addition, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the method described in the aforementioned method embodiment are implemented.

[0074] The present invention also provides an electronic device, comprising:

[0075] one or more processors; and

[0076] A memory associated with the one or more processors, the memory is used to store program instructions, and when the program instructions are read and executed by the one or more processors, the steps of the method described in the method embodiment of the above-mentioned embodiment 1 are executed.

[0077] It should be noted that a part of the present invention can be applied as a computer program product, such as a computer program instruction, when it is executed by an intelligent electronic device (such as a smart mobile phone or a tablet computer, etc.), the method and / or technical solution according to the present invention can be called or provided through the operation of the intelligent electronic device. The program instruction for calling the method of the present invention may be stored in a fixed or removable recording medium, and / or transmitted through a data stream in a broadcast or other signal-bearing medium, and / or stored in a working memory of an intelligent electronic device that operates according to the program instruction. Here, according to an embodiment of the present invention, a device is included, the device including a memory for storing computer program instructions and a processor for executing program instructions, wherein when the computer program instruction is executed by the processor, the device is triggered to run the method and / or technical solution based on the aforementioned multiple embodiments of the present invention.

[0078] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which aspect, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present invention is defined by the appended claims rather than the description of the above embodiments, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A voice wake-up interactive response method, comprising the following steps: Get user's speech; Detecting whether the user's speech contains a wake-up word; After the wake-up word is detected, it is detected whether there is a user speaking in the detection area after the wake-up word; Wherein, the detection within the detection area includes detecting whether a user speaks within a normal speaking detection area; If no user speech is detected in the detection area, a welcome message is played; The detection within the detection area also includes: Detecting whether a user speaks in the false trigger area in the time area immediately after the wake-up word is detected, and if a speech is detected, further determining whether the speech is confused with the tail sound of the wake-up word; if there is confusion, it is considered that no speech is detected; and / or The blind spot detection area behind the normal speech detection area detects whether the user speaks. If no speech content is detected in the blind spot detection area, the fundamental frequency is calculated for the blind spot detection area. If the fundamental frequency calculation can obtain speech with a certain duration, it is considered that the speech is detected.

2. The method according to claim 1, characterized in that: The wake-up model is used to detect the user speaking in the detection area.

3. The method according to claim 1, characterized in that A correlation analysis method is used to determine whether the utterance is confused with the tail sound of the wake-up word.

4. A voice wake-up interactive response system, the system comprising: The acquisition module is used to obtain the user's speech; A wake-up word detection module, used to detect whether there is a wake-up word in the user speech acquired by the acquisition unit; A user speech detection module is used to detect the second speech of the user after the wake-up word; Wherein, the user speech detection module includes a normal speech detection module, and includes a false trigger detection module and / or a blind spot detection module; The false trigger detection module is used to detect whether the second utterance exists within a period of time immediately after the wake-up word, and if the second utterance is detected, it is necessary to further determine whether the second utterance is confused with the tail sound of the wake-up word; The normal speech detection module is used to detect whether the second speech exists within a period of time after the wake-up word or the false trigger detection; The blind spot detection module is used to detect whether the second speech exists within a period of time after the normal speech detection, and if the second speech is still not detected, perform baseband detection in the time period to determine whether the user speaks; The voice output module is used to output a welcome message when the user speech detection module does not detect the second speech after the system wakes up.

5. The system according to claim 4, characterized in that The user speech detection module uses a wake-up model to perform detection when detecting the second speech.

6. The system according to claim 4, characterized in that The false trigger detection module uses a correlation analysis method to determine whether the second utterance is confused with the tail sound of the wake-up word.

7. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of implementing the method according to any one of claims 1 to 3 are implemented.

8. An electronic device comprising: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method according to any one of claims 1 to 3.

Images